Meet the analysts helping IPL teams decide on strategy
- The IPL paved the way for a huge amount of data to be collected into a database for analysts; now teams constantly rely on sophisticated number crunching.
We all know how Virat Kohli loves to play in the ‘V’. He is equally good at probing acute angles behind square, especially in the shorter versions where he has made a name for himself as one of the most successful chase masters ever. But what is his get-out-of jail strategy, the first shot that he almost always resorts to at the beginning? More often than not, it’s towards the leg-side.
So, at one time, bowlers from an opponent team had a clear brief on what to do with him: bowl 18 inches outside his off stump for the first four overs. Cricket is increasingly turning towards technology to help decipher patterns—strengths, weaknesses and quirks—captured through hundreds of hours of footage, for modern day analysts to dig deep and come up with information that may provide a tactical edge.
It’s not an exact science; but the way international teams and every franchise in the IPL has been leaning more and more on data analysis proves its undeniable importance. Mumbai Indians, the most successful team in the IPL, has a dedicated performance data app that can be accessed year-round by the current players and management of the franchise. “The game is getting smarter. Athletes are getting smarter. People who are managing athletes are getting smarter. From a strategy perspective, it makes the game very exciting,” said CKM Dhananjai, Mumbai Indians’ data performance manager.
Some stuff can be as basic, like the boundary conversion metric of Chris Gayle who remains a hot pick even at 41 because 76% of his total IPL runs have come in boundaries and sixes, much ahead of Kohli, Rohit Sharma, David Warner or AB de Villiers. Other metrics can be infinitesimally focused—like a batsman who’s comfortable playing short balls at or under 135kmph, but falters when the ball is delivered at around 140kmph. Or, the probable reason why Rajasthan Royals bought Chris Morris at an exorbitant price despite the South African not having a great IPL record—data from Cricviz reveals that no one has induced more false shots during Powerplays in the last three seasons than Morris (27.7%). Jofra Archer comes second on that list with 27.2%. But Archer has an economy rate of 4.5 in that same period, compared to Morris’s 7.65. If you were an analyst sifting through this data, you may come up with this conclusion: Morris is extremely hard to handle at the critical Powerplay stage, but he is unlucky; since he has this enviable skill, his luck is bound to turn.
The volume of raw data that cricket produces can be dizzying. Dhananjai estimates that the franchises in the IPL together track nearly a thousand players across the world. “Each team has 25 players, multiply that into eight—that’s your first 200 players who need to be tracked. For player recruitment/auctions, you have to track 5x players. That tracking happens over a 24-36 month window,” Dhananjai said. For each player being tracked, an enormous amount of ball-by-ball data is generated. “Initially (first 2-3 years of the IPL, where analysts focused on building databases) we were collecting 20-30 pieces of information per ball,” said M Lakshmi Narayanan, analyst at Chennai Super Kings. “Now I’m capturing 120-130 parameters per ball. For example a batsman is comfortable playing a 135km bouncer but not at 140km. We piece together data on how uncomfortable a batsman is against the ball coming in or going out. Some batsmen don’t score much in the first 10 balls but his strike rate changes in the next 10-20 balls. What exactly does he do to facilitate that?”
Given the variables cricket accommodates, the parameters can be tweaked any way analysts want. Analysts now mine data on venues, length of boundaries, pitches, types of soil, kind of breeze (think Fremantle Doctor at Perth, Hooghly at Eden, Arabian Sea at Wankhede or Indian Ocean at Chennai) and even how the height of a bowler affects trajectories. It’s the analysts’ job then, to root out what’s relevant for a player, team, or situation. “We don’t give so much data that the player gets confused. Say we have 10 patterns but out of those, I will give data pertaining to only two,” said Lakshmi.
Professional data analysis in cricket is still a relatively new phenomenon. India’s first full-time analyst was Subramaniam Ramakrishnan, popularly known as Ramky, who joined the Indian team when John Wright arrived from Kent to become the head coach in 2000. At first, the analysts worked with what is now so ubiquitious that we now call it “basic data”, like, is a particular left arm fast bowler more effective against left-handed batsmen or right-handed batsmen? Now we have 20 years worth of data and a quick algorithm to tell us that in a few seconds—just head over to Espncricinfo.com’s Statsguru. Back then, Ramky had to do create his own database and make his own calculations.
Then, the BCCI began to provide match footage from untelevised domestic games for analysis—now Ramky and others with similar interests could collect ball-by-ball data—for that same left-arm bowler for example, they could now generate trajectories, release points, speeds, etc. In 2006, Ramky founded SportsMechanics, a company specialising in sports data analysis, working off the database he had so painstakingly created. “First there was only match analysis based on non-televised video data,” said Lakshmi. “Slowly, after 1-2 years, once we got that data, we started to map it realistically.” The next big explosion of data access happened with the IPL.
“When I was at ESPN Star Sports (where he started the popular cricket fantasy series Super Selector), we had started getting some more data in terms of where the ball pitched and what it did,” recounted Joy Bhattacharjya, one of India’s earliest data evangelists and sports producers who was also Kolkata Knight Riders’ team director till 2014. “But the data wasn’t too rich. Once the IPL started (in 2008) it gave us a lot of data in a very short time.” That data was critical not just for figuring out game plans, but also for making the right decisions at the IPL auctions. “I had worked with a bunch of scientists from ISI Delhi to design a programme for buying in the auction. We used it in 2008,” said Bhattacharjya.
“In 2014, we worked with SAP (a multinational software company) who was one of our partners to create a decision support system for the squad that we bought.” On the international level, England were already familiarising their players with Hawk-Eye data. Nathan Leamon, analyst with the England team since 2009, set up that system. Last December, Leamon experimented with a signal system during a T20 in Cape Town, placing placards with numbers and letters on team balcony for England captain Eoin Morgan to decode and implement. With an eye on the T20 World Cup in October, Leamon is now working with Eoin Morgan at Kolkata Knight Riders this season.
So how does cricket data compare against those available in baseball, basketball or football? “(In cricket) every ball is what we call a discrete incident. Every ball produces 25-30 pieces of data,” said Bhattacharjya. “Football isn’t that data rich. Where we are not close to what is being done internationally is biomechanical data. Compared to NBA or football, we are still scratching the surface.”